Phase Roadmap

All 16 phases are approved. Phases 1-11 are deployed and live. Phases 12-16 are scaffolded (types, stores, service stubs, panel UIs, Cloud Function stubs) but their ML/compute backends are not yet connected.

Phase Status

Phase	Feature	Status	Key Files
1	Auth (email/password + Google OAuth)	Live	AuthPage, useAuthStore, auth.ts
2	Landing page with WebGL hero	Live	yugma-landing/
3	Scene templates + loader	Live	sceneTemplates.ts, ProjectInfoPanel
4	Share links (URL-encoded)	Live	shareLinkService, urlEncoder
5	Library panel (primitives + GLTF + Sketchfab)	Live	LibraryPanel/
6	Real-time collaboration (room codes, cursors)	Live	collabService, CollabOverlay
7	AI Scene Composer	Live	AIPanel, aiCompose, aiSerializer
8	Text-to-3D (Meshy)	Live	GenerateSection, generationService
9	Advanced collab (comments, feed)	Live	CommentPin, ChangeFeed
10	Export (GLB, screenshot, embed)	Live	exportUtils, EmbedPage
11	AI Materials (30 presets, AI gen)	Live	materialPresets, AITextureSection
12	Video-to-3D reconstruction	Scaffolded	video.types, useVideoStore, VideoPanel
13	Industrial digital twins	Scaffolded + Mounted	twin.types, useTwinStore, TwinPanel
14	Product vibe-coder	Scaffolded	product.types, useProductStore, ProductPanel
15	Factory simulation (physics)	Scaffolded + Wrapped	physics.types, usePhysicsStore, PhysicsWorld
16	AI cinematic director	Scaffolded	cinematic.types, useCinematicStore, CinematicPanel

AI Enhancements (Cross-Phase)

These enhancements improve the AI layer across all phases:

Enhancement	Status	Files
Spatial preprocessor (circle/grid/stack/spiral/line/scatter)	Complete	spatialPreprocessor.ts
Cross-session memory (Firestore)	Complete	aiService.ts (loadLastSession/saveAISession)
Style fingerprint + memory	Complete	styleFingerprint.ts
Planner→executor decomposition (3D-GPT)	Complete	aiCompose.ts (PLANNING_SYSTEM_PROMPT)
Real focus_camera (smooth tween)	Complete	CameraController.tsx, useSceneStore
YSL v1.5 schema (userData, semanticRole, relationships)	Complete	scene.types.ts, aiSerializer.ts

What Each Scaffolded Phase Needs to Go Live

Phase 12 — Video-to-3D

Build a Cloud Run orchestrator that chains: SAM2 segmentation → monocular depth estimation → object classification → auto-rigging → scene composition. The Cloud Function startVideoReconstruction creates the job doc; the orchestrator processes it and writes back resultGlbUrl.

Phase 13 — Digital Twins

1. Deploy RTDB rules for /sensors/* (firebase deploy --only database)
2. Build an MQTT→RTDB bridge (Cloud Run or Cloud Function) that receives sensor readings and writes to sensors/{id}/latest
3. The useSensorSubscriptions hook + pushReading → updateObject pipeline is already wired end-to-end

Phase 14 — Product Vibe-Coder

1. Set NEXAR_API_KEY secret for the Octopart/Nexar GraphQL API
2. Replace the mock data in octopartProxy.ts with real API calls
3. Wire enclosure geometry into SceneRenderer (convert Enclosure → box with mounting hole subtractions)

Phase 15 — Factory Simulation

1. pnpm --filter yugma-app add @react-three/rapier @dimforge/rapier3d-compat
2. Replace PhysicsWorld.tsx stub body with <Physics paused={!running}>{children}</Physics>
3. Add RigidBody wrappers per SceneObject based on usePhysicsStore.bodies

Phase 16 — AI Cinematic Director

1. Build renderStoryboard Cloud Function: headless R3F render → ffmpeg → video upload
2. Wire CINEMATIC_DIRECTOR_PROMPT into a new Cloud Function that takes a scene + brief and returns a Storyboard JSON
3. Build a shot timeline player in CinematicPanel that drives setCameraTarget per shot